CN116824870B

CN116824870B - Road segment flow prediction method, device, equipment and storage medium

Info

Publication number: CN116824870B
Application number: CN202311111666.XA
Authority: CN
Inventors: 蔡明祥; 徐昊; 李程; 曾萼岚; 王译; 罗滨锋; 胡雨铭
Original assignee: Guojiao Space Information Technology Beijing Co ltd
Current assignee: Guojiao Space Information Technology Beijing Co ltd
Priority date: 2023-08-31
Filing date: 2023-08-31
Publication date: 2023-11-24
Anticipated expiration: 2043-08-31
Also published as: CN116824870A

Abstract

The embodiment of the disclosure provides a road section flow prediction method, a road section flow prediction device, road section flow prediction equipment and a storage medium, and is applied to the technical field of big data traffic management. The method comprises the steps of obtaining driving data of a vehicle; the driving data are obtained by the edge computing node according to the road network data and GPS point position data generated by the movement of the vehicle; obtaining road traffic flow based on the driving data and a preset time window; constructing a traffic flow augmentation matrix according to the road traffic flow and a preset pavement attribute quantization value; performing convolution mechanism processing on the traffic flow augmentation matrix to obtain a traffic flow input sequence; and inputting the traffic flow input sequence into a pre-constructed road section flow prediction model, and outputting a traffic flow condition prediction result. In this way, the cleaning and the extraction of the GPS track data of the vehicle are completed at the edge end, the storage and processing pressure of the data center platform is reduced, the road surface attribute characteristics are endowed to the traffic flow, and the accuracy of traffic flow prediction is improved.

Description

Road segment flow prediction method, device, equipment and storage medium

Technical Field

The disclosure relates to the technical field of big data traffic management, in particular to a road section flow prediction method, a road section flow prediction device, road section flow prediction equipment and a storage medium.

Background

Along with the continuous improvement of urban digitization degree, the data volume acquired by the traffic management department is continuously increased, so that great influence is brought to data cleaning, application and storage, and a technical means for improving the calculation efficiency and reducing the storage cost is needed to reduce the calculation and storage cost of the traffic management department. Meanwhile, the existing traffic flow prediction algorithm does not consider the road surface condition corresponding to the vehicle driving road section, and is difficult to excavate the influence of different road grades and materials on the traffic flow. The method specifically comprises the following aspects:

(1) The calculation and storage pressure of the traditional vehicle GPS track data processing mode is high. The traditional vehicle GPS track data is generally stored and managed in a centralized way by a data center of a traffic management department, along with the continuous increase of the number of vehicles, the data amount required to be stored by the data center is rapidly increased, the data center generally stores all vehicle track data in a centralized way, the stored data is required to be cleaned when the GPS data is classified and cleaned, the calculation and storage cost of the GPS track data is increased, and the increasing data analysis requirement is difficult to meet.

(2) The existing vehicle track extraction algorithm has low extraction performance and is easy to lose track details. The existing vehicle track extraction algorithm is represented by a track-placian-placken algorithm with GPS (global positioning system) thinning, the core thought is to connect a starting point with an end point, keep the point farthest from the line, record the point to be saved after multiple iterations, and finally obtain the extracted track data. Furthermore, such methods still require at least two nodes to be saved while minimizing the number of recorded nodes. With the continuous development of urban road networks, the road networks are more complex, more track features are lost when the track extraction is performed by using the method, and the algorithm iteration times are increased, so that the problems of extraction performance reduction and the like are caused.

(3) The existing traffic flow prediction algorithm focuses on traffic flow, and the influence of road surface properties on the traffic flow is ignored. The current traffic flow prediction algorithm mainly takes traffic flow as a research object, generally starts from two dimensions of time and space, researches the change rule of the traffic flow along with the time and the space, and the most typical algorithm is a diffusion convolution recurrent neural network DCRNN, a graph convolution network T-GCN and other methods. At the same time, some studies have considered external factors other than traffic flow, such as: weather, holidays, etc. However, the influence of the road surface attribute on the traffic flow is not considered, and in the display condition, the possible flow of roads which are the same national roads is larger, and the possible flow of road sections with more lanes is also larger, so that the potential relation between the traffic flow and the road surface attribute is difficult to extract by the conventional traffic flow prediction algorithm, and the problem of low prediction accuracy is caused. And secondly, the existing traffic flow prediction algorithm predicts the road section nodes in multiple directions, and fails to pay attention to the traffic flow condition of the whole road section in a certain time.

Disclosure of Invention

The disclosure provides a road section flow prediction method, a road section flow prediction device, road section flow prediction equipment and a storage medium.

According to a first aspect of the present disclosure, a road segment traffic prediction method is provided. The method comprises the following steps:

acquiring running data of a vehicle; the driving data are obtained by the edge computing node according to the road network data and GPS point position data generated by the movement of the vehicle; the edge computing node is a vehicle-mounted GPS;

obtaining road traffic flow based on the driving data and a preset time window; the road traffic flow comprises traffic flows corresponding to a plurality of time points;

constructing a traffic flow augmentation matrix according to the road traffic flow and a preset pavement attribute quantization value;

performing convolution mechanism processing on the traffic flow augmentation matrix to obtain a traffic flow input sequence; wherein the traffic flow input sequence comprises a plurality of traffic flow sequences with equal lengths;

and inputting the traffic flow input sequence into a pre-constructed road section flow prediction model, and outputting a traffic flow condition prediction result.

Further, obtaining driving data according to the road network data and GPS point position data generated by vehicle movement, including:

The edge computing node acquires road network data and GPS point position data of a target vehicle; the road network data are road network data in a vehicle-mounted GPS of the target vehicle; the road network data comprises a road section ID and road section data; the GPS point location data is GPS point location data generated by the movement of the target vehicle; the GPS point location data comprises a plurality of GPS point locations;

the edge computing node computes the distance between the GPS point location and the road section data, and determines the current road section ID of the target vehicle to obtain a target driving road section ID;

the edge computing node records the running time of the target vehicle on a target running road section and generates a running record field; the driving record field comprises the time of a target vehicle driving into a target driving road section, the time of the target vehicle driving out of the target driving road section and the driving time of the target vehicle in the target driving road section;

the edge calculation node transmits the travel record field and the target travel section ID as travel data of the target vehicle to the data center server.

Further, the pavement property quantized values include a road grade quantized value and a road material quantized value;

The constructing a traffic flow augmentation matrix according to the road traffic flow and the preset pavement attribute quantized value comprises the following steps:

，

where Z represents the constructed traffic flow augmentation matrix,represents the traffic flow of road segment i at point in time t,representing a road grade quantized value corresponding to the road section i; />And the quantized value of the road material corresponding to the road section i is represented.

Further, the processing the traffic flow augmentation matrix by a convolution mechanism to obtain a traffic flow input sequence includes:

and compressing the dimension of the traffic flow augmentation matrix Z through a convolution mechanism, and segmenting the dimension into a plurality of sequences with equal length to obtain a traffic flow input sequence.

Further, the process of constructing the road section flow prediction model includes:

constructing a road section flow prediction model based on the TCN and the spatial attention mechanism; wherein, the TCN is TCN with a gating mechanism.

Further, the method further comprises:

setting colors on the road traffic flow in sections according to the flow, and constructing an urban road network thermodynamic diagram;

and generating a road congestion degree live map according to the urban road network thermodynamic diagram so as to facilitate traffic management personnel to dynamically adjust traffic operation management strategies according to the road congestion degree live map.

According to a second aspect of the present disclosure, a road segment flow prediction apparatus is provided. The device comprises:

the data acquisition module is used for acquiring the driving data of the vehicle; the driving data are obtained by the edge computing node according to the road network data and GPS point position data generated by the movement of the vehicle; the edge computing node is a vehicle-mounted GPS;

the traffic flow generation module is used for obtaining road traffic flow based on the running data and a preset time window; the road traffic flow comprises traffic flows corresponding to a plurality of time points;

the matrix construction module is used for constructing a traffic flow augmentation matrix according to the road traffic flow and the preset pavement attribute quantized value;

the input sequence generation module is used for carrying out convolution mechanism processing on the traffic flow augmentation matrix to obtain a traffic flow input sequence; wherein the traffic flow input sequence comprises a plurality of traffic flow sequences with equal lengths;

and the prediction result generation module is used for inputting the traffic flow input sequence into a pre-constructed road section flow prediction model and outputting a traffic flow condition prediction result.

According to a third aspect of the present disclosure, an electronic device is provided. The electronic device includes: a memory and a processor, the memory having stored thereon a computer program, the processor implementing the method as described above when executing the program.

According to a fourth aspect of the present disclosure, there is provided a computer readable storage medium having stored thereon a computer program which when executed by a processor implements a method according to the first aspect of the present disclosure.

The embodiment of the disclosure provides a road section flow prediction method, a device, equipment and a storage medium, which utilize an edge calculation frame to complete the cleaning and extraction of vehicle GPS track data at an edge end, reduce the storage and processing pressure of a data center station, then construct a traffic flow augmentation matrix fused with road surface attributes, endow road surface attribute characteristics for traffic flows, extract the time correlation inside traffic flows of all road sections of a city in parallel through a time convolution network TCN on the basis, extract the space correlation among all road sections by taking a traffic flow sequence corresponding to each road section as a unit in combination with a space attention mechanism, finally obtain the traffic flow condition of all road sections of the whole city road network in future time, and improve the accuracy of traffic flow prediction.

It should be understood that what is described in this summary is not intended to limit the critical or essential features of the embodiments of the disclosure nor to limit the scope of the disclosure. Other features of the present disclosure will become apparent from the following description.

Drawings

The above and other features, advantages and aspects of embodiments of the present disclosure will become more apparent by reference to the following detailed description when taken in conjunction with the accompanying drawings. For a better understanding of the present disclosure, and without limiting the disclosure thereto, the same or similar reference numerals denote the same or similar elements, wherein:

FIG. 1 illustrates an architectural diagram of road segment traffic prediction according to an embodiment of the present disclosure;

FIG. 2 illustrates a schematic diagram of GPS trajectory acquisition and road segment traffic prediction, according to an embodiment of the present disclosure;

fig. 3 illustrates a flowchart of a road segment traffic prediction method according to an embodiment of the present disclosure;

FIG. 4 shows a schematic diagram of a process of generating travel data according to an embodiment of the disclosure;

fig. 5 shows a block diagram of a road segment flow prediction apparatus according to an embodiment of the present disclosure;

fig. 6 illustrates a block diagram of an exemplary electronic device capable of implementing embodiments of the present disclosure.

Detailed Description

For the purposes of making the objects, technical solutions and advantages of the embodiments of the present disclosure more apparent, the technical solutions of the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present disclosure, and it is apparent that the described embodiments are some embodiments of the present disclosure, but not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments in this disclosure without inventive faculty, are intended to be within the scope of this disclosure.

In addition, the term "and/or" herein is merely an association relationship describing an association object, and means that three relationships may exist, for example, a and/or B may mean: a exists alone, A and B exist together, and B exists alone. In addition, the character "/" herein generally indicates that the front and rear associated objects are an "or" relationship.

The disclosure relates to a GPS track acquisition and road section flow prediction method based on edge calculation and TCN, which is a schematic diagram of road section flow prediction architecture shown in FIG. 1, and comprises an edge calculation layer, a data storage layer and a flow prediction layer. The edge computing layer comprises a plurality of edge computing nodes taking a vehicle-mounted GPS as a unit, a GPS track cleaning and extracting algorithm is deployed on the basis, GPS track data are processed by using vehicle self equipment, reasonable allocation of computing resources is facilitated, and a processing result is sent to the data storage layer, namely the data center server, so that the storage pressure of the data storage layer is reduced. The data storage layer mainly comprises a data storage medium for storing traffic flow and road surface numbers and an urban traffic network thermodynamic diagram constructed based on traffic flow data, wherein the traffic flow is stored by taking road sections as units, and important traffic point thermodynamic diagrams are generated by searching and self-adapting segmentation of traffic flow conditions. The traffic flow prediction layer constructs a traffic flow augmentation matrix fused with road surface characteristics according to the data stored in the data storage layer, and predicts the future traffic flow by using a deep learning model based on TCN and a spatial attention mechanism.

The road segment flow prediction method of the present disclosure is generally described below with reference to fig. 2.

As shown in the schematic diagram of GPS track collection and road traffic prediction in fig. 2, firstly, edge computing nodes are constructed, vehicle driving tracks are associated with road networks, a GPS track cleaning and extracting algorithm is deployed, after drift track correction is performed, driving tracks are extracted after vehicles drive out of road sections; then transmitting the data to a traffic flow data center station, constructing a traffic flow data storage structure taking a road section as a unit, and generating a traffic flow network thermodynamic diagram based on traffic flow and a self-adaptive segmentation algorithm; secondly, constructing a traffic flow augmentation matrix integrating road surface properties and a traffic flow prediction model based on TCN and a spatial attention mechanism in the road surface information data, and finally completing prediction of traffic flow of each road section of the city for a period of time in the future.

Fig. 3 shows a flow chart of a road segment flow prediction method 300 according to an embodiment of the present disclosure. Applied to a data center server, the method 300 includes:

in step 310, travel data of the vehicle is acquired.

The driving data are obtained by the edge computing node according to the road network data and GPS point position data generated by the movement of the vehicle; the edge computing node is a vehicle-mounted GPS.

In some embodiments, the schematic diagram of the process of generating the driving data shown in fig. 4 is applied to an edge computing node, that is, the driving data is obtained according to the road network data and the GPS point location data generated by the movement of the vehicle, and includes the following steps:

in step 410, the edge computing node obtains road network data and GPS point location data of the target vehicle.

The road network data are road network data in a vehicle-mounted GPS of the target vehicle; the road network data comprises a road section ID and road section data; the GPS point location data is GPS point location data generated by the movement of the target vehicle; the GPS point location data includes a plurality of GPS point locations.

And step 420, the edge computing node computes the distance between the GPS point location and the road segment data, determines the road segment ID of the current target vehicle running, and obtains the target running road segment ID.

In step 430, the edge computing node records the running time of the target vehicle on the target running road section, and generates a running record field.

The driving record field includes a time when the target vehicle is driven into the target driving section, a time when the target vehicle is driven out of the target driving section, and a driving time of the target vehicle on the target driving section.

In step 440, the edge computing node transmits the travel record field and the target travel section ID as travel data of the target vehicle to the data center server.

In some embodiments, participants in urban traffic include freight cars, passenger cars, private cars, etc., which are typically equipped with onboard GPS. The existing data storage and processing mode generally comprises the steps of uniformly uploading vehicle GPS data to a data center server, cleaning and calculating original data when the vehicle GPS track data are required to be analyzed, and increasing the storage pressure of the data center server along with the rapid increase of the number of vehicles. The existing vehicle track cleaning and extracting algorithm generally needs to be iterated for a plurality of times, and meanwhile the problems of track feature loss and the like exist. Therefore, the method is based on an edge computing frame, takes a vehicle-mounted GPS as an edge computing node, and deploys a GPS track cleaning and extracting algorithm based on combination of vehicle GPS points and road sections, so that the vehicle extracts the running track of the vehicle in the running process and then sends the running track to the middle platform, storage and processing pressure of a data middle platform server is reduced, and allocation of computing resources is further optimized. Meanwhile, a compressed extraction part of the vehicle running track is deployed on a vehicle-mounted GPS sensor, and the running track is directly extracted in the GPS sensor to obtain running data.

In some embodiments, a vehicle-mounted GPS is used as an edge computing node, and the distance between a point location and a road segment data is computed by using the road network data in the vehicle-mounted GPS and GPS point location data generated by the movement of the vehicle to determine the road segment on which the current vehicle is traveling, which is specifically as follows:

，

wherein the method comprises the steps ofRoad segment ID representing the nearest road segment corresponding to point location i, < >>Representing a set of road networks->Representing a set of points i>A calculation function representing the point and line geodesic distances.

In some embodiments, correction of the travel track is required to prevent road segment identification errors caused by GPS drift due to weather, buildings, etc. Assuming that there are 10 GPS points for continuous travel,solving for the corresponding pointA sequence of travel sections of the vehicle can then be obtained, in particular as follows:

，

if there are a plurality of points running on the same road segment, from which one to two points suddenly leave, but which in the following points are returned again to the original road segment, the points leaving this road segment are considered to be error points, which need to be corrected, in particular if the sequence satisfiesAt the same timeIt is considered that the (i+7) th and (i+8) th points may be due to GPS point drift caused by weather factors, and the two points should be corrected at this time, which is specifically as follows:

，

After correction, the 10 GPS points of the continuous travel can be regarded as the case of traveling on the same road section. If a plurality of continuous points run on a road section, a part of subsequent points have already run off the road section, and the plurality of continuous points do not return to the original road section, the GPS point of the vehicle is considered to be normal at the moment, correction is not needed, and meanwhile, the running of a car on the previous road section is also explained to be finished. At this time, the running record of the vehicle on the previous road section is compressed through the point location thinning algorithm, and a plurality of running points on the same road section can be replaced by one point location and other data entries. The point location thinning mainly comprises two steps: firstly, calculating the central point of the driving point of the same road section, wherein the specific mode is as follows:

，

wherein the method comprises the steps ofRepresents the distance between point i and point j, < +.>Representing a set of points j ∈>The sum of the distances between the point i and the same road segment and n nodes is represented by +.>Indicated is the cluster center point, i.e. the point with the smallest sum of the distances from other points among the points in the same road segment driving. After solving the central point, performing a second step: recording the time of the vehicle entering the road section, exiting the road section and passing through the road section, and generating a driving record field, thereby recording the whole process of the vehicle entering and exiting the road section, specifically comprising the following steps:

，

Wherein the method comprises the steps ofIndicating that the corresponding vehicle is in road section +>Time of passage of>Representing the road segment ID->Indicating the time of departure +.>The time of entry is indicated. After the vehicle enters a new road section, the edge computing node cleans and extracts the track according to the running data of the previous road section, finally generates a field and transmits the field to the data center server, the running track on the road section is represented by a plurality of fields, the light storage of the running track of the vehicle is realized, the details of the track of the vehicle are ensured, the technical support is provided for the GPS track extraction of the vehicle in various scenes, and the field structure is as follows:

，

and 320, obtaining the road traffic flow based on the driving data and a preset time window.

The road traffic flow comprises traffic flows corresponding to a plurality of time points.

In some embodiments, the vehicle driving data extracted in step 440 may form a driving record of each vehicle on each road section of the city, where a fixed time window is set to extract the traffic flow on the road section by taking the road section as a unit, so as to determine whether the vehicle is driving on the road section within a specified period of time as an evaluation criterion of the road section flow. Suppose there are 10 trolleys at 17:00 driving-in road sections a,17:20 travel out section a,5 dollies at 17:10 entry road sections a,17:30 out of road segment a, and the set time window length is 10 minutes, 17:00 to 17: the flow rate of the 10 road section a is 10, 17:10 to 17: the flow rate of the 20 road section a is 15, 17:20 to 17: the flow rate of the 30 segment a is 5. Based on the extraction mode, a time sequence database storage structure taking time as a main key can be constructed, which is equivalent to slicing the traffic flow of the city according to time, the road traffic flow corresponding to each time point is stored as a storage sequence, and the data recorded in one row in the database are specifically as follows:

，

Wherein the method comprises the steps ofRespectively represent road sections->Is set to be +.>Corresponding traffic flow->And so on … …. By the data extraction mode, a data foundation is laid for generating the urban road network thermodynamic diagram and constructing the traffic flow prediction model. Meanwhile, the constructed GPS track cleaning and extracting algorithm can dilute the running track on the same road section into a point position, and the key information of the track is reserved by adding data fields, so that the high-level thinning of the GPS track is realized, and the GPS track is combined with a road network to form a complete vehicle running track.

And 330, constructing a traffic flow augmentation matrix according to the road traffic flow and the preset pavement property quantized value.

In some embodiments, a traffic flow augmentation matrix may be constructed that blends road surface attribute features based on extracted road traffic flow and road surface data, exerting an impact on the road surface attributes on the traffic flow. The specific implementation mode is as follows:

traffic flow is usually closely related to road segment grades, road segment materials and the like, so road surface factors such as the road segment grades are fused to predict future traffic flow, and road segment characteristics are fused for flow prediction. The pavement attribute is generally composed of soft data such as national roads, provinces and counties, cement, asphalt, sand and the like, and the data cannot be directly used for data analysis work, so that the soft data needs to be quantized in the following specific quantization modes:

，

Wherein the method comprises the steps ofRepresenting road section->Corresponding road class quantization result,/->Representation->And (5) a corresponding quantification result of the road material. Besides the soft data to be quantized, the data such as the number of lanes and the like which are not required to be quantized can be used as a part of the traffic flow augmentation matrix, and the traffic flow augmentation matrix can be constructed according to the quantized structure, and the traffic flow augmentation matrix is specifically as follows:

，

where Z represents the constructed traffic flow augmentation matrix,represents the traffic flow of road segment i at point in time t,quantized value representing road class corresponding to road segment i, < ->And the quantized value of the road material corresponding to the road section i is represented.

And 340, performing convolution mechanism processing on the traffic flow augmentation matrix to obtain a traffic flow input sequence.

Wherein the traffic flow input sequence comprises a plurality of equal length traffic flow sequences.

In some embodiments, to reduce the calculation amount of the subsequent model, the dimension of the augmentation matrix Z is compressed by a convolution mechanism, and the specific calculation manner is as follows:

，

where F represents the result after convolution, and is also the input to the deep learning model,the convolution operation is represented, segmentation is carried out, the convolution operation is divided into a plurality of input sequences with equal length, and the input sequences are fused with the road surface attribute based on the steps, so that the initial data of the input model can be obtained. The dimension of the traffic flow augmentation matrix Z is compressed through a convolution mechanism and is segmented into a plurality of sequences with equal length, and a traffic flow input sequence is obtained.

And 350, inputting the traffic flow input sequence into a pre-constructed road section flow prediction model, and outputting a traffic flow condition prediction result.

In some embodiments, traffic flow is closely related to temporal and spatial factors, so future traffic flow needs to be predicted from both temporal and spatial perspectives. The process for constructing the road section flow prediction model comprises the following steps: constructing a road section flow prediction model based on the TCN and the spatial attention mechanism; because the urban road network comprises a plurality of road sections and the adjacent road sections inevitably have mutual influence, a spatial attention mechanism is embedded in the TCN, the capturing capacity of the prediction model in terms of spatial correlation is improved, and the prediction accuracy is further improved. The TCN is a TCN with a gating mechanism, a spatial attention mechanism is applied to each TCN in the stacking process of multiple layers of TCNs, and spatial correlation existing between input traffic of each road section is further extracted, so that the view angle of the road section is wider, road information is fused, and prediction is more accurate. In order to extract the time correlation in the traffic flow, a TCN structure with a gating mechanism is constructed for extraction, and the specific calculation mode is as follows:

，

Wherein the method comprises the steps ofRepresenting a feature with a temporal dependence, wherein +.>Representing diffusion convolution,/->Hadamard product representing matrix, +.>Representing the sigmoid function, i.e. the gating mechanism, the main function of which is to filter potentially noisy data in the sequence,/->Representing an activation function->Input sequence representing input layer μ TCN, +.>And->Respectively represent the learnable parameters of the first and second convolution kernels corresponding to the layer mu TCN mechanism.

On the basis, the present disclosure is also characterized in thatA spatial attention mechanism is applied, and potential correlation before each sequence is considered mainly from the sequence corresponding to each road section, so that a query matrix, a key matrix and a value matrix corresponding to each road section are needed to be calculated firstly, and the method is concretely as follows:

，

wherein the method comprises the steps of、/>、/>Respectively express characteristic->A corresponding query matrix, key matrix and value matrix,the self-adaptive matrixes corresponding to the three matrixes are respectively represented, and the spatial attention mechanism focuses on the relation among sequences, so that the self-adaptive matrixes are different from the common channel attention mechanism, and the self-adaptive matrixes are specifically as follows:

，

wherein the method comprises the steps ofRepresenting one vector in the query matrix corresponding to one node, namely one vector in the query matrix corresponding to the ith node, and a plurality of shapes like +. >Vectors of (2) can constitute->Similarly, the->、/>And can also be calculated according to the mode. />The feature after the time correlation is extracted from the sequence corresponding to the road section i is shown. Thereby ensuring that the spatial attention mechanism is implemented in the basic unit of a node sequence. The process of calculating spatial attention is specifically as follows:

，

wherein the method comprises the steps ofRepresenting the result obtained after spatial attention mechanism extraction, T represents the transpose operation, < >>Representing the dimension of the input data ∈>Representing vectors in the query matrix +.>Representing vectors in the key matrix,/->Representing the vector in the value matrix, softmax represents the normalized exponential function. On the basis, in order to reduce the feature loss caused by the continuous deepening of the depth of the neural network, residual connection is added after each calculation to relieve the problem of the feature loss, and the result after the residual connection is used as the input of the next TCN structure, specifically as follows:

，

along with the change of the extracted characteristics of the increasing number of layers of the TCN+ space attention structure, an intermediate storage unit is designed for recording the data processed each time, and the final prediction result is obtained from the final potential characteristics under different convolution depths of a connection mechanism and a full connection layer, wherein the specific implementation mode is as follows:

，

Wherein the method comprises the steps ofRepresenting data recorded in the intermediate storage unit +.>The connection operation is specifically indicated by transversely splicing the two matrices, and by the term +>The spatial attention mechanism extraction result of the h-th layer TCN is shown, and h is set according to the sequence length. />Representing traffic flow conditions for the future time period t+1 to 2t,/o>Full connectivity layer operation is shown.

Based on the foregoing embodiment, the method of still another embodiment provided in the present disclosure further includes:

In some embodiments, based on the road traffic flow obtained in step 320, traffic flows can be counted according to different apertures, and a thermodynamic diagram is drawn according to the size of the traffic flow, while from the overall point of view, the road segments with larger traffic flows are relatively fewer, and most road segments with smaller traffic flows, from the point of view of the traffic flow distribution of the city, the data is L-shaped, and the data is difficult to reasonably divide into sections by using the traditional segmentation means, so that the original traffic flow data sequence is divided into multiple sections by an adaptive segmentation algorithm. The method comprises the following steps:

Firstly, sorting according to the flow, dividing the flow into k equal parts according to the data volume, wherein each part of data volume is the same, the k size can be set by itself, and at the moment, the whole urban road network thermodynamic diagram is divided into even k equal parts, but the road sections with larger traffic flow are difficult to excavate, the road section with the largest urban traffic flow is 50000, the road section with the tenth traffic flow rank is 2000, and the urban roads are thousands of, and the two road sections are likely to be divided together. Based on the thought that the change in the same interval should be as small as possible, on the basis of the first division interval, the data in each interval is adaptively adjusted, specifically by calculating a threshold value based on the classified interval s after classificationAnd with a set acceptable thresholdComparison is made, if there is->Dividing the interval into two sections and calculating the threshold value respectively until all the intervals meetStopping the iteration at the time, wherein->The specific calculation mode of (2) is as follows:

，

wherein the method comprises the steps ofRepresenting standard deviation of data in interval s，/>Representing the mean of the data in interval s. The reasonable segmentation setting based on the traffic flow can be realized based on the steps, the construction of the urban road network thermodynamic diagram is finally realized, the urban road network thermodynamic diagram can be used as a reference for identifying the road congestion degree and dynamically adjusting by a traffic management department, and meanwhile, from the aspect of the overall thermal condition, the reference can be provided for making a maintenance plan for a road maintenance department, a series of decision support platforms and systems can be expanded, and the like.

It should be noted that, for simplicity of description, the foregoing method embodiments are all described as a series of acts, but it should be understood by those skilled in the art that the present disclosure is not limited by the order of acts described, as some steps may be performed in other orders or concurrently in accordance with the present disclosure. Further, those skilled in the art will also appreciate that the embodiments described in the specification are all alternative embodiments, and that the acts and modules referred to are not necessarily required by the present disclosure.

The foregoing is a description of embodiments of the method, and the following further describes embodiments of the present disclosure through examples of apparatus.

Fig. 5 shows a block diagram of a road segment traffic prediction apparatus 500 according to an embodiment of the present disclosure, applied to a data center station server. As shown in fig. 5, the apparatus 500 includes:

a data acquisition module 510 for acquiring driving data of the vehicle; the driving data are obtained by the edge computing node according to the road network data and GPS point position data generated by the movement of the vehicle; the edge computing node is a vehicle-mounted GPS;

the traffic flow generating module 520 is configured to obtain a road traffic flow based on the driving data and a preset time window; the road traffic flow comprises traffic flows corresponding to a plurality of time points;

A matrix construction module 530, configured to construct a traffic flow augmentation matrix according to the road traffic flow and a preset pavement attribute quantization value;

an input sequence generating module 540, configured to perform convolution mechanism processing on the traffic flow augmentation matrix to obtain a traffic flow input sequence; wherein the traffic flow input sequence comprises a plurality of traffic flow sequences with equal lengths;

the prediction result generating module 550 is configured to input the traffic flow input sequence into a pre-constructed road segment flow prediction model, and output a traffic flow condition prediction result.

It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the described modules may refer to corresponding procedures in the foregoing method embodiments, which are not described herein again.

According to an embodiment of the disclosure, the disclosure further provides an electronic device, a readable storage medium.

Fig. 6 shows a schematic block diagram of an electronic device 600 that may be used to implement embodiments of the present disclosure. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the disclosure described and/or claimed herein.

The electronic device 600 includes a computing unit 601 that can perform various appropriate actions and processes according to a computer program stored in a ROM602 or a computer program loaded from a storage unit 608 into a RAM 603. In the RAM603, various programs and data required for the operation of the electronic device 600 can also be stored. The computing unit 601, ROM602, and RAM603 are connected to each other by a bus 604. An I/O interface 605 is also connected to bus 604.

A number of components in the electronic device 600 are connected to the I/O interface 605, including: an input unit 606 such as a keyboard, mouse, etc.; an output unit 607 such as various types of displays, speakers, and the like; a storage unit 608, such as a magnetic disk, optical disk, or the like; and a communication unit 609 such as a network card, modem, wireless communication transceiver, etc. The communication unit 609 allows the electronic device 600 to exchange information/data with other devices through a computer network, such as the internet, and/or various telecommunication networks.

The computing unit 601 may be a variety of general and/or special purpose processing components having processing and computing capabilities. Some examples of computing unit 601 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized Artificial Intelligence (AI) computing chips, various computing units running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as method 300. For example, in some embodiments, the method 300 may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as the storage unit 608. In some embodiments, part or all of the computer program may be loaded and/or installed onto the electronic device 600 via the ROM602 and/or the communication unit 609. One or more of the steps of the method 300 described above may be performed when a computer program is loaded into RAM603 and executed by the computing unit 601. Alternatively, in other embodiments, computing unit 601 may be configured to perform method 300 by any other suitable means (e.g., by means of firmware).

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuit systems, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), systems On Chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs, the one or more computer programs may be executed and/or interpreted on a programmable system including at least one programmable processor, which may be a special purpose or general-purpose programmable processor, that may receive data and instructions from, and transmit data and instructions to, a storage system, at least one input device, and at least one output device.

Program code for carrying out methods of the present disclosure may be written in any combination of one or more programming languages. These program code may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus such that the program code, when executed by the processor or controller, causes the functions/operations specified in the flowchart and/or block diagram to be implemented. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and pointing device (e.g., a mouse or trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic input, speech input, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a background component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such background, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), wide Area Networks (WANs), and the internet.

The computer system may include a client and a server. The client and server are typically remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server may be a cloud server, a server of a distributed system, or a server incorporating a blockchain.

It should be appreciated that various forms of the flows shown above may be used to reorder, add, or delete steps. For example, the steps recited in the present disclosure may be performed in parallel, sequentially, or in a different order, provided that the desired results of the disclosed aspects are achieved, and are not limited herein.

The above detailed description should not be taken as limiting the scope of the present disclosure. It will be apparent to those skilled in the art that various modifications, combinations, sub-combinations and alternatives are possible, depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present disclosure are intended to be included within the scope of the present disclosure.

Claims

1. The road section flow prediction method is characterized by being applied to a data center server and comprising the following steps of:

Constructing a traffic flow augmentation matrix according to the road traffic flow and a preset pavement attribute quantization value; the pavement attribute quantized values comprise road grade quantized values and road material quantized values;

the constructed traffic flow augmentation matrix is as follows:

；

where Z represents the constructed traffic flow augmentation matrix,representing the traffic flow of road section i at time t, < >>Representing a road grade quantized value corresponding to the road section i; />Representing road segment i pairA corresponding quantized value of road material;

inputting the traffic flow input sequence into a pre-constructed road section flow prediction model, and outputting a traffic flow condition prediction result; the road section flow prediction model is constructed based on TCN and a spatial attention mechanism.

2. The method of claim 1, wherein deriving travel data from road network data and GPS point location data generated by vehicle movement comprises:

3. The method of claim 1, wherein the convolving the traffic flow augmentation matrix to obtain a traffic flow input sequence comprises:

4. The method of claim 1, wherein the TCN is a TCN with a gating mechanism.

5. The method according to claim 1, wherein the method further comprises:

6. A road segment traffic prediction apparatus, applied to a data center server, comprising:

the matrix construction module is used for constructing a traffic flow augmentation matrix according to the road traffic flow and the preset pavement attribute quantized value; the pavement attribute quantized values comprise road grade quantized values and road material quantized values; the constructed traffic flow augmentation matrix is as follows:

；

Where Z represents the constructed traffic flow augmentation matrix,representing the traffic flow of road section i at time t, < >>Representing a road grade quantized value corresponding to the road section i; />Representing a quantized value of a road material corresponding to the road section i;

the prediction result generation module is used for inputting the traffic flow input sequence into a pre-constructed road section flow prediction model and outputting a traffic flow condition prediction result; the road section flow prediction model is constructed based on TCN and a spatial attention mechanism.

7. An electronic device, comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-5.

8. A non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the method of any one of claims 1-5.